JPS6085372A - Method and device for determining mean frequency by time region system - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Background of the Invention] The present invention provides a method for estimating the average frequency of a signal that has good performance over a wide range of signal-to-noise ratios and is performed in real time. Regarding equipment.

Accurately determining the average frequency of a time-varying signal is desirable, for example, in Doppler suite ultrasound velocity measurements where the average frequency of the signal corresponds to the average velocity of the flow field being sampled.

Other applications include frequency and phase modulation communication systems and speech recognition. Acoustic measurement of blood velocity is based on the Doppler effect. Perhaps the greatest chore in such measurements is accurately determining the Doppler frequency gradient I shift in a noisy environment. We provide a powerful time domain method for determining the Doppler frequency shift such that t=-V, which satisfies this strict condition.

FIG. 1 shows one conventional method of time domain processing using the I10 algorithm. This is Doppler I (in phase) and Q
It is so named because it extracts the average frequency directly from the (orthogonal) signal. Ultrasonic Imaging, Vol. 2, pp. 232-261 (July 1980)
See the article "Energy Spectral Center Point Detector for Doppler '4A" by Nori, Gerzberg and J. D. Meindl, published in 1993. When the S/N ratio is high and the frequency is low, this method can give results comparable to the Fourier transform method! However, there are two main drawbacks. The resulting average frequency varies according to the sine of the true mean frequency, which causes erroneous readings on platforms with large frequency excursions. 11.1 When there is noise in the signal, the noise power is divided by iJi (P-12-1 of equation (7) described later)
-02), but is it a pheasant in the molecule? Blue is smoothed. Due to the noise term in the denominator, the S/N ratio is 1.
When approaching , a large error occurs. The Fourier variation method has a similar form and estimates the noise power by 1J to obtain a reasonable estimate of the average frequency.

SUMMARY OF THE INVENTION In one particular embodiment of the invention, an improved time-domain method for determining average frequency demodulates a time-varying signal using a quadrature-phase criterion and It includes simultaneously filtering and sampling the demodulated signal. Both the I and 0 time samples are extended over an integer number of periods, typically one period. Multiply the non-delayed time sample by the non-delayed time sample as the numerator and the minute B1, both of which are subject to noise. Create an uncorrelated difference signal and a sum signal.The denominator is the product of each sample and the sample shifted by 1 for one cycle, and the sum (+j: 'j is
' / with a low-frequency wave and a backward averaging method (backward average method).
running avcrage process) to substantially remove extraneous sounds in both terms. The smoothed difference signal and 31! From the arc tangent of the ratio of the collapsed sum signal, the time-varying average frequency of No. 18 is derived. Performance suffers because of the time shift in the signal correlation function.

With this device and method, it is possible to obtain an accurate value of tFtlt'' + iE of the frequency, and to obtain a result of the frequency domain Fourier transform method and the ratio W-1 when the signal-to-noise ratio is good!lr. Much better results are obtained for signal-to-noise ratios lower than the human OdB, which cannot be used.The average frequency is linear with respect to the average power of the spectrum.Feature #3 of the algorithm of the present invention is to automatically compensate for frequencies higher than the Nyquist frequency J by 1Ili when determining the average frequency.

Improved ultrasound methods and pulsed Doppler devices for measuring velocity in blood and similar fluids are detailed, and other applications are also discussed. The received echoes Gj are suitably coherently demodulated to baseband and the focused in-phase and quadrature Doppler signals are C-gated in relation to range at the repetition frequency of the transmitted J pulses. The average frequency of the Doppler signal corresponds to the frequency deviation.

The average blood velocity is determined and displayed as a function of time.

DETAILED DESCRIPTION OF THE INVENTION The basic Doppler equation describes the frequency deviation (Δ
f), ■ of the entrance wave is j wave number (fo), sound speed in the propagation medium (C), and angle between the direction of movement and the direction of sound propagation (θ
°) and is expressed as follows.

Rewriting this, the Doppler velocity is increased by 4!7 as follows.

In a typical application, C is known, to is conveniently chosen, and θ can be measured. Therefore, determining Δf corresponds to knowing the speed of scattering i 41.

There are a number of ways to extract particle velocity information from this equation.

In this section, we will discuss the phase coherence pulsed Doppler system. (See FIG. 5 for a similar device that is part of a duplex imaging format with a common transducer for both the imaging modality and the Doppler modality). The measured waveform is a sample of the waveform relative to the transmission time.
It consists of 1 and Q Doppler (g) sampled simultaneously at a constant distance determined by the delay of the pulse and at a pulse repetition frequency f. The purpose of processing the Doppler signal is to extract particles or particles from this measurement signal. The purpose is to extract the velocity information of the red head.

Since the measured velocity is related to the frequency deviation, it is natural to process the signal in the frequency domain.

However, it has the advantage of being an interview-like time-domain method. Specifically, the sampled Doppler signals I and Q can be viewed as the real and imaginary parts of a complex time domain waveform. The instantaneous frequency of this waveform corresponds to the Doppler shift Δt in equation (1). Since the time-domain waveform is sampled at the pulse repetition frequency, the highest resolvable frequency is determined by the Nyquist criterion that two samples per recycle is the minimum that characterizes a band-limited signal. is limited to fr/2. However, since both ■ and Q signals can be used, the sign of the shift is known, and therefore the equation (
2) determines the direction of the velocity of the scatterer (target).

Representing a time-domain signal by two time-domain signals that are quadrature in phase is common in communication systems where the two signals are extracted during the demodulation process.

f(t)−I(t)+iQ(t) (3) The algorithm for determining the average frequency is derived as described below, based on frequency modulation and phase modulation communication systems, voice recognition, Doppler ultrasonic velocity measurement, It can also be applied to other uses depending on the case. The average frequency of this signal (1) G; L, the power spectrum of the signal S(ω〉) is generally expressed as:
By taking the Fourier transform f of the signal and squaring the width of this transform, we get -(1!j. (The asterisk represents a complex conjugate number.) △ △ S (ω) = f * f (5) Using Plancherel's theorem and the Fourier transform formula, the average frequency can also be expressed as a time integral instead of an integral over frequency.The equivalent time integral form is as follows.

If the signal is expressed in the form of 1 and Q in equation (3), the average frequency of this equation is as follows.

This is the I10 algorithm mentioned earlier. In the pulsed Doppler system, the time signal is a data signal that is sampled discretely rather than continuously.

Attempts to determine the average frequency from the difference equation in equation (7) were unsuccessful. There is a power term in the denominator, so the noise has a correlation, and as a result, as the S/N ratio approaches 1, the error increases. Other shortcomings have also been identified.

The hardware representing an apparatus for directly determining the average of a spectrum using this method is shown in FIG. Frequency-modulated continuous time-varying signal F (t
) are fed to two product demodulators 10.11. The standards for these demodulators are 90° relative to each other to remove the carrier.
°There is a phase shift. The demodulated signals 1(t) and Q(t) are passed through a low-pass filter 12.13 (symbols in the figure indicate that center and high frequencies are rejected) to restore the envelope. and is supplied to multipliers 14 and 15. Differentials are obtained in circuits 16 and 17 and supplied to a multiplier. The output of the multiplier is the cross term shown in the numerator of equation (7). circuit 18
The cross term is subtracted at , and the result is sent to a low-pass filter 19 for integration. The demodulated in-phase signal and quadrature signal are squared by multipliers 20 and 21, and the sum of the squares is calculated by circuit 22, as shown in the denominator of equation (7). The unit 23 integrates. The ratio obtained by divider 24'C is the average frequency ω.

Instead, if expressed in a compound form, f(t)−A(t)exp(iθ(t))
(8) In this case, equation (7) becomes as follows.

The instantaneous phase expressed by ■ and Q is The instantaneous frequency ω can be determined by the phase differential dθ/'dt.

Differentiating Equation (10) and using A2=12+02=P(t)=P (11), we find that (I) (t) is the instantaneous power) Equations (7) and (9) are the same. It is easy to show that there is.

In the present invention, equations (9) and (10) are implemented in such a way that the average frequency of the signal can be determined by one. This implementation is based on the recognition of mathematical sophistication that has been overlooked until now. Using this equation as a starting point, the average frequency of the sampled signal is. where t is the time between samples (assumed to be uniform), i is the number of samples, N+1 is the total number of simples, and further. J-superior performance is obtained by recognizing that Δθ1 can be expanded in analytical form instead of simply taking the difference between the two calculated phases. At this time, the difference will be as follows.

or It ] + − I + to h Qf −+ The numerator of this last equation is the same as when implementing equation (7) using the finite difference method, assuming that the samples are of the same finite time interval. However, the denominator is quite different.

The denominator has the product of each time sample and the time ripple reshifted by one period. This is the first extension term of the autocorrelation function, not the power. As a result, the noise of both terms (the numerator has a cross term) can be smoothed independently. If the sample delay time is longer than the noise correlation time, the noise is smoothed out and therefore reduced relative to the signal. This condition is actually achieved by the capacity (16)
When implementing, it is preferable to take advantage of the fact that the instantaneous power 1)(t) is relatively constant over time to determine the average frequency.

In other words, the following approximation Δθ1 must−−(17) Δt is valid, and the numerator (N) and denominator (
D) can be used in a smoothed form.

However, N,=ku1-α)N+-+ +α(It Q=+Qt
It-+ ) <19) Di=(1-α) D+-+
+α(It It-+ 4-Ch Qt-+)'
(20) The factor α is smaller than 1 and is called the optimal value, but it depends on the application. Typical values for Doppler measurements are 1/
It is 16 to 1/256. Formulas (19) and (20)
The running average of is used in place of the sum that is passed to other equations. Two important characteristics of the signal make S
/N ratio can be significantly improved.

First, the terms 1 + Qt-+ -Qt I i-1 and l1li-10 + Q +-+ vary slowly with respect to time in most practical applications. The rate of change is much smaller than the instantaneous frequency of I or Q alone. Therefore, the frequency response of the J, S\:, position is smooth and does not have a noticeable chromatic effect. Second, the terms in the denominator form a cross-correlation function of d3 over one delay period. In the case of uncorrelated noise, if the smoothing time is increased, the noise should converge to the bald spot.

FIG. 2 is a block diagram of signal processing for determining the spectral average using the instantaneous frequency algorithms of equations (18), (19), and (20).

Although this applies to continuous time-domain signals and sampled time-domain signals, the latter signal will be illustrated and described. Demodulator 10', 1 with quadrature reference
The low 1ilX filters 12', 13' for filtering out higher harmonics of the 1' and residual carrier frequencies and their fractions may be the same as in FIG. It is. The first step in this detection process is typically to demodulate the signal to remove the carrier. Optimal performance is obtained by using synchronous detection. Demodulation can be done directly to baseband (homodyne)
Or it may be via an intermediate frequency (hederodyne). P-wave demodulated in-phase and quadrature signals 1(t) and (,)(1
) is sampled at a selected rate in sampling circuits 25, 26 and digitized at this point. Individual machines [IL can be digital, analog, or digital/
Analog hybridization can also be implemented using any current or future technology. The main element is delay operation τ
, and this is due to the temporal change of the signal correlation function,
Improve performance. the in-phase and quadrature time samples in an integer number of periods), preferably by one period; the time delay device 2;
7.28 will extend R. The delayed in-phase ripple and the R-undeferred quadrature samples are sent to multiplier 29. Similarly, the delayed quadrature samples and the undelayed in-phase lean pull are sent to multiplier 30. This first cross term is subtracted from the second cross term in circuit 31 to produce a difference signal uncorrelated to noise. The delayed and undelayed in-phase time samples are sent to a third multiplier 32, the Masanobu and undelayed quadrature time samples are sent to a fourth multiplier 33, and the products are summed in an adder 34. do,
Generates a sum signal with no correlation to noise. The denominator consists of the product of each time sample and the time sample shifted by one period, and therefore the -C noise is uncorrelated.

The numerator difference signal and the denominator sum signal are independently and separately low-pass filtered to smooth the C1 noise and substantially remove noise from both. Equations (19) and (20) derive smoothed terms in the numerator and denominator by the backward averaging method. First, the old term in the numerator is multiplied by (1-α) and added to the new term multiplied by α. Factor α is typically 1/16 to 1/2
56, but the number of time samples to be averaged is 16 to 2
56.

The smoothed and substantially noise-free difference and sum signals are divided by a divider 37, and by determining the arctangent of this ratio, the average frequency 5 of the output signal, i.e. the time-varying signal, is determined. generated. Determining the arctangent can be done at a much slower rate than sampling.

An example of the effect of the instantaneous frequency algorithm compared to the average frequency calculated by the Fourier transform is shown in Figures 3a-3e. Curve F is the result obtained by the Fourier transform method, and the average noise power has been corrected. Curve - ``is α = 1/64 and formula (18)
It is based on a time domain algorithm. For comparison,
The curve "is shifted by 25 units. Each figure is from experimental measurements with different signal-to-noise ratios (S/N ratios), which are 24 dB and 4 dB, respectively.
, OdB, -4dB and -8dB L, , IC6S
When the N ratio is O dB (Fig. 3a), the peak of curve F is still discernible, but below O dB, the Fourier transform method cannot be used and in fact the last curve F (the third
Figure c) is a straight line. In contrast, the time-domain method of the present invention provides considerable performance even at a signal-to-shoulder ratio of -4 dB, -8 dB, while important features of the velocity curve can still be seen. The main advantage of this method is that the time-domain algorithm can extract average frequency and velocity information even in noisy environments.

The second feature of this algorithm is that the calculated average frequency is linear with the average power of the spectrum. This is shown in FIG.

In Fig. 4, there are two frequencies ω1 and ω2, and the total power is between ω2 and ω1, as β-PwJ/Pt, and the ratio β
changes linearly. The average frequency must be −βω1+(1−
It can be seen that it is expressed as β)ω2.

The third important feature of this algorithm is that it can automatically compensate for frequencies higher than the Nyquist frequency ωN when calculating the average frequency. This holds true as long as the entire spectrum of the instantaneous signal centered around the average frequency lies within an interval whose width is smaller than the Nyquist interval. This helps determine the average frequency when aliasing (aliasin(1)) occurs.

In this invention, the average frequency of a frequency-modulated time-varying signal, whether in digital or analog form, is
Alternatively, a digital/analog hybrid method can be obtained by using a hardware circuit or a computer.

In a digital hardware arrangement, in FIG. 2, the delay time can be determined by one or more clock cycle delay devices, and it is advantageous to use two or more time delay devices. In analog H6, this delay is obtained by a delay line. In the sampling data device, COD
, (charge-coupled device) type devices, packet brigates and similar devices can be used to implement the -U,W extension, as well as to perform some signal processing. The configuration of the device is an Intel 8231 to achieve most, if not all, of these functions with sufficient storage.
A special processing chip such as △ can be used. By testing the sign and relative magnitude of each term in the numerator and denominator, the arctangent can be determined by identifying angular intervals that are within 45 degrees. Once the arguments within 45° are known, a piecewise linear approximation can be used.

The duplex imaging device of FIG. 5 is a real-time, single fan-directed beam scanner that incorporates the time-domain Doppler processor of the present invention for measuring blood flow velocity in real time. . This will be briefly explained, but is described in detail in US Pat. No. 4,217,909 and US Pat. No. 4,155,260. However, it is recognized that the pulsed Doppler device can be constructed as a separate device and need not be part of a duplex device with B-scan imaging capabilities.

The illustrated device has a common linear transducer array 39 made up of a number of transducer elements 40, which transducer elements are energized by a pulse driver @ 41 to form an ultrasound beam 42. and transmits ultrasonic pulses. During the Doppler mode of operation, the beam is directed at a scanning angle and distance 1i1
The transmitted ultrasound pulse impinges on a selected sample volume 43 at tR and measuring the blood flow velocity therein. The required %fi degree of the delay time is significantly reduced, while maintaining the precision of the phase focusing action that can be achieved more easily.
The main components of the receive channel, which features baseband signal processing to achieve good lateral resolution, are a wideband receiver 44 for all channels, a balanced demodulator and low-pass filter 45, and a time delay device. 46 and an adder 47. Each receive channel has parallel I (in-phase) and Q
(orthogonal) processing channels in which the received error signals are electronically directed and dynamically focused. Using a phase quadrature emission frequency reference, the echo signals are amplified and demodulated, and the output of each demodulator is low-frequency waveformed, leaving an envelope that is then delayed. If the path lengths are different enough, add a delay proportional to the path length difference before coherent addition. In the imaging mode of operation, the added and focused signals are applied to circuit 48 to create a composite signal.

This composite signal is sent to the cathode ray tube 49 to form a fan-shaped image.
This is the video signal sent to.

In the Doppler mode of operation, the instrument's controller 50 causes an RF pulse of ultrasound to be transmitted and impinge on the sample volume 43.
is set by the user as Jζ, and the distance gate 51 is used to sample from a desired depth and detect the velocity pattern at a specific location. Another feature of exciting this transducer is that it allows the backscattered echoes from fast and slow moving blood cells within the rimple volume to be properly sampled at various distances. , the repetition period is variable. 4KH2, 8K
Several settings for the ultrasonic pulse repetition frequency are available, such as Hz and 16Kl-1z.

The added (τ) and focused 2 and Q signals are directly provided to the Doppler device t8 without generating a composite signal thereof. The focused in-phase and quadrature signals ΣI and ΣQ are sampled at a specified time after each transducer excitation start, corresponding to the distance (and the time it takes for the ultrasound signal to return to the transducer). Range gate 51 is listened to by controller 50 for a relatively short period of time at times corresponding to receiving backscattered echoes from the sample volume, extracting a pair of analog time samples in parallel. It can be seen that moxibustion is determined by the pulse repetition frequency. The in-phase and quadrature time samples are digitized and sent to a time-domain Doppler processor 52 where an arctangent algorithm is used to derive the average frequency of the time-varying Doppler signal. After this, the blood velocity is obtained from equation (2).

This ultrasonic device has a Doppler-style display for two types of speed information. First, the Doppler frequency is displayed on the oscilloscope 53 so that the user can observe its changes in real time. A strip chart recording device 54 shows the evolution of velocity versus time.
- A hard copy is printed. Either the averaged average velocity or the entire distribution of velocity (positive and negative) can be displayed as the blood flow approaches or moves away from the transducer array. multiplexed ECG
The signal provides a time reference for events that occur during the course of a cardiac cycle.

The time-domain Doppler processor 52 is preferably implemented in a Birdows configuration, like most real-time frequency-domain Fourier processors. In order to test such a pulsed Doppler device, the algorithm was constructed using Rafter 1A and a time domain analysis of C1 was calculated. The program calculated the moving average of the numerator and denominator of the arctangent argument. After determining the arctangent of the ratio, the calculated phase change with respect to time was compared to the cutoff frequency of a high pass filter designed to reject direct current and low frequency components. If it is within the stopband of the filter, a simple I10 approximation is used to estimate the phase change. Once the phase change is calculated, it is multiplied by a conversion factor to give the velocity. The current values of I and Q are stored to form a new value of delay μ.

The ease of determining the instantaneous mean frequency using this method allows the use of other signal processing features related to spectral line estimation. Part 1
The second is a time-frequency histogram. In many applications, estimation of the actual spectrum is desired. This is usually done using a Fourier transform. By extending the time domain method, an approximate spectral distribution can be obtained by forming a histogram of instantaneous power relative to instantaneous frequency. The second feature is
This concerns the correction of unbalance of the duplexer (1) and Q (14). In Doppler and other quadrature demodulators, the final 41 degree of precision of the device is affected by the accuracy of the balance of the 1 and Q demodulators. A new scheme for multiplexing detectors has been developed.

Other applications of this method for determining the average frequency of time-varying signals include frequency modulation and phase modulation communication systems and speech recognition. In this communication system, an average frequency circuit is used as an FM detector. The output obtained from this restores the modulated signal of the 1j0 transmission. In speech recognition, the measured average frequency is the information (speech) signal applied to a carrier wave and represents the average pitch of the speech. Audio 1-
The rack is sampled and digitized, a time domain algorithm is used to extract the average frequency, and then the audio bits are medicated.

In summary, the superior performance and ease of implementation of this time-domain scheme is advantageous for time-domain signal processing. This device has the advantage over classical Fourier transform schemes of better performance at lower cost, as well as better ability to cope with noisy environments.

This invention will be specifically illustrated with reference to preferred embodiments.
However, it goes without saying that those skilled in the art can make various modifications within the scope of the present invention.

[Brief explanation of the drawing]

Figure 1 shows conventional signal processing for determining the average value of a spectrum using the I10 algorithm.
Figure 2 is a block diagram of a device implementing an improved method for determining the average frequency in the time domain using an instantaneous frequency algorithm, and Figures 3a to 3e are (a)
S/N-24 dB, (b) S/N=4 dB, (
c) S/N=OdB. (d) S/N=-4dB and (0) S/N
--Graph showing pulsed velocity measurements using time-domain and frequency-domain signal processing for various signal-to-noise ratios of 8d13; Figure 4 shows different relative powers but constant total power. FIG. 5 is a simplified block diagram of a duplex ultrasound imaging system incorporating the time-domain Doppler processing system of the present invention. (Explanation of main codes) 10', 11': Demodulator, 12', 13', 35.36: Low-frequency single wave generator, 2
5, 26: Sampling circuit, 27, 28: Delay circuit, 29, 30.32.33: 11) Multiplier, 31: Subtractor, 34: Adder, 37: Divider, 38: Arctangent circuit. Saki 61 (Ze) Time (erect time C rot)

Claims (1)

[Claims] 1) In a method of determining the average frequency using a signal-to-noise ratio above and below OdB using a time domain method, a time-varying signal is demodulated and r-wave demodulated in-phase and quadrature generate a signal,
Both these signals are simultaneously delayed by the same extent and multiplied by the undelayed and delayed in-phase and quadrature signals to create uncorrelated difference and sum signal terms of the aE tone, respectively, and the difference and sum signals are The time period is determined by subtracting the terms separately to substantially remove the noise in both terms and determining the arctangent of the ratio of the smoothed and relatively noiseless difference and sum signals. a method comprising the step of obtaining an output signal representative of the average frequency of a signal that varies over time. 2. In the method described in claim 1), the demodulated signal is generated by homodyne demodulation to the baseband and a low frequency wave for removing the residual carrier frequency and its higher harmonics. How to do it. multiply by the delayed quadrature signal and the undelayed in-phase signal, subtract the products to produce the difference signal, and multiply by the delayed and undelayed in-phase signals;
A method of multiplying the delayed and non-delayed quadrature signals and consolidating their products to form the sum signal. 4) Demodulating the time-varying signal using a quadrature reference in a method for determining the average frequency of a frequency-modulated time-varying signal having a wide range of signal-to-noise ratios in a time-domain manner. sample the demodulated signal at a selected period to produce in-phase and quadrature samples, multiply the in-phase and quadrature samples by an integer number of periods or delays, and multiply the delayed in-phase samples by undelayed quadrature samples; The delayed quadrature sample is multiplied by the undelayed in-phase ripple and the respective products are subtracted to generate the difference (M), multiplied by the delayed and undelayed in-phase ripple and the delayed and delayed Multiply the quadrature samples without correlation and add the products to create a sum signal that is uncorrelated with the noise.
low-pass filtering the difference signal and the sum signal separately to smooth and substantially reduce the noise in both, and dividing the difference signal by the sum signal;
A method comprising the step of deriving the average frequency of said time-varying signal from the arctangent of the thus formed. 5) A method according to claim 4) in which - C1 said time samples are all delayed by one period. 6) A method according to claim 4, in which the difference signal and the sum signal are filtered and smoothed by backward averaging. 7) In a method of measuring the velocity of the flow of blood and similar liquids using ultrasound, an ultrasound pulse is generated to scan a selected sample volume, the echo signals are received, and the phase is quadrature. Demodulating the echo signal using an emission frequency reference and processing it to generate a focused in-phase and quadrature Doppler signal, distance gating the Doppler signal after each pulse transmission, the period of which is determined by the pulse repetition frequency. extracting in-phase and quadrature samples, extending the in-phase and quadrature samples for at least one period, multiplying the in-phase samples by the quadrature samples without delay, and multiplying the quadrature samples by the missing in-phase samples; The respective products are subtracted to form a difference signal sample, multiplied by the delayed and undelayed in-phase time samples, multiplied by the delayed /j and undelayed quadrature time samples, and the products are added to eliminate the noise. Create uncorrelated sum signal samples, low-pass filter the difference signal and the sum signal separately to smooth and substantially reduce the noise in both, and calculate the ratio of the smoothed difference signal to the smoothed sum signal. A method comprising the steps of deriving the average frequency of the Doppler signal, and hence the C frequency deviation, from the arctangent of C, determining the average blood velocity, and displaying the velocity as a function of time. 8) A method according to claim 7, in which the difference signal and the sum signal are filtered and smoothed by backward averaging. 9) A method for determining Doppler frequency shift and blood velocity in real time in the method according to claim 8). 10) In an apparatus for estimating an average frequency using a time domain method, means for demodulating a temporally varying signal and low-pass filtering it to generate a demodulated in-phase and quadrature signal, and a means for demodulating the signal at a selected frequency. means for simultaneously converting the signals to extract in-phase and quadrature temporal randomization; and means for determining the average frequency of the temporally varying signal at a signal-to-noise ratio above and below OdB; The average wave number determining means is means for delaying the time ripple, means for multiplying the sample without delay and the delayed sample to generate a difference signal and a sum signal, respectively, which are uncorrelated with noise;
Apparatus including means for separately smoothing the difference signal and the sum signal to substantially remove uncorrelated noise, and means for taking the arctangent of the ratio of the smoothed difference signal to the smoothed sum signal. 11) The apparatus according to claim 10, wherein the delaying means simultaneously delays the time samples by one period between samples, and the smoothing means comprises a moving average -
A device consisting of a rice bowl. 12) In an ultrasonic device for measuring the velocity of blood and similar liquids, a means for transmitting ultrasonic pulses and receiving echoes to scan a selected sample volume, and an emission frequency orthogonal in phase. means for demodulating the echo signal to baseband using a reference and focusing and summing to generate an in-phase and quadrature time-varying Doppler signal; gating means for simultaneously gating the signal to extract in-phase and quadrature time samples 1 and Q representing echoes backscattered from the sample volume and generating delayed time samples 11-1 and Q t −+; At the same time, a good signal-to-noise ratio and a good signal-to-noise ratio can be achieved by processing the undelayed samples 11 and Ql and the delayed samples so as to smooth uncorrelated noise and eliminate almost all of it. means for determining, in real time, the average frequency, and therefore the frequency deviation, of said time-varying signal in real time; and means for determining the average blood velocity and displaying the velocity as a function of time. 13) In the ultrasonic device according to claim 12), the means for determining the average frequency is characterized in that Δ°C is a period between samples, i is a number of time samples, and N is a number of time samples. The numerator, D is the denominator, α is a number smaller than 1, and N1 = (1-α) N + + α (1: Qt- + Qt
1+-+) DI=(1-α) D knee t+α(1:
It−+ +Q+ Qt−4) and an ultrasound device that implements a −0th order time domain algorithm. 14) In the ultrasonic device according to claim 13), the gating means extracts time samples at a pulse repetition frequency, and the means for determining the average frequency extracts the samples only during IFJI during lean pull. An ultrasound device including means for delaying.